Etching Out Influence: Exploring ASML as a Strategic Asset in EU-US Relations
By Alina Hueber (Spring ‘25 Talos Alumni) and Antoine De Greef
Note: This piece was drafted as part of a policy brief project during the Talos Fellowship. The views and opinions expressed in this policy paper are those of the author(s) and do not necessarily reflect the official policy or position of Talos Network.
1 Abstract
ASML’s monopoly in EUV lithography is an underutilised strategic asset, offering significant opportunities to protect EU interests. This brief investigates whether the EU could leverage ASML to advance AI safety objectives and secure a meaningful share of the AI value they’re helping create. It also touches on using ASML to counter economic coercion through strategic export controls under the Anti-Coercion Instrument, pointing out that this is likely not the best use for this asset. Further, the brief identifies six significant challenges to implementation and explores how some could be mitigated long-term. The brief highlights that, while any strategy leveraging ASML would likely be costly, the costs may be acceptable in scenarios where the stakes are sufficiently high. The option space to leverage ASML on an EU level remains severely understudied. Rather than offering definitive recommendations, this brief serves to open up discussion by identifying possible policy approaches with the associated implementation challenges, tradeoffs and areas requiring further research.
2 Strategic Context
The Dutch firm ASML holds a monopoly on Extreme Ultraviolet (EUV) lithography machines, which are essential for manufacturing advanced AI chips at nodes below 7nm. Very roughly, ASML builds machines used by companies like TSMC to build AI chips designed by companies like Nvidia. These chips provide computing power (compute), which is likely to remain one of the three necessary pillars of AI development alongside algorithms and data. Research suggests that among these pillars, compute makes by far the biggest contribution to AI progress - its impact is estimated to be twice as great as that of algorithmic advancements. In AI training, lower-node chips offer huge performance advantages, since they not only allow more transistors to be packed into the same area but are also adept at AI-specific computations. Because AI models are scaling exponentially in size, the need for compute grows by orders of magnitude — and only cutting-edge chips made with the smallest nodes can keep up.
Despite this crucial contribution, the EU has little control over how safely AI is developed and how downstream profits are distributed, since frontier AI development is strongly focused in the U.S. and China. Given the potential for cross-border catastrophic risks from frontier models — ranging from misuse to misalignment or loss of human oversight — the EU may be compelled to act once progress outpaces safety frameworks and foreign actors push ahead regardless of societal consequences. This urgency is heightened by the U.S.-China race for AI leadership, which prioritizes capability gains over safety investments. The precise moment when the EU should step in is hard to pin down given the sliding nature of AI progress, but setting a risk threshold may be useful. The EU can leverage its standard-setting power and provide a democratic counterweight to the growing autocratic nature of the U.S. and China.
Even without catastrophic risks materializing, AI is poised to reshape society, with some estimating a potential impact akin to the industrial revolution. To protect its interests in an AI-dominated world, the EU may want to ensure it captures a fair share of profits from the AI deployment enabled by their technology.
The EU also faces a more transactional international environment, highlighted by recent U.S. tariffs on European goods. The U.S. has previously leveraged its position in the semiconductor supply chain to advance its strategic interests through export controls. This raises the question whether the European Union could leverage ASML via the Anti-Coercion Instrument to protect its interests.
To explore these possibilities, we highlight three venues for further research, including AI safety licensing requirements, an export tax or royalty system and the use of European-level export controls on advanced EUV lithography machines. We focus on measures targeting the U.S. and exclude China from this discussion - partly to limit the scope of this piece and partly because EUV export controls towards China are already in place and less neglected. We identify six implementation challenges and explore options for mitigation, but recognize the need for further study. The option space to leverage ASML on an EU level is severely understudied.
3 Policy Approaches
3.1 AI Safety Licensing Requirements
As the sole provider of the machines necessary to the fabrication of advanced AI chips, ASML is in the unique position to limit the speed of scaling. After all, a higher fabrication capacity for chips means faster scaling for data centers. Given the high risks that AI is likely to pose, this might be an invaluable lever for AI Safety.
Just as the U.S. has restricted access to AI chips to trusted entities through the AI Diffusion Framework and their Foreign Direct Product Rule (FDPR), the EU could develop similar mechanisms to ensure that products created using EU technology don’t cause spill-over costs for the EU. Even though the new Trump administration has recently abolished the AI Diffusion Framework, they still use the FDPR to impose export restrictions on both TSMC and ASML on the grounds that their products are made using U.S. technology. An EU equivalent to the FDPR could ensure that certain AI chips made using EUV lithography machines require export licences by the EU. By this logic, the EU could require e.g. Nvidia to limit their AI chip sales to AI developers and cloud providers committed to addressing shared AI risks. Just as the U.S. bans certain countries and companies that act against their interest through their Entity List, the EU could blacklist companies that do not follow certain AI safety requirements — or even hold a licence from an AI safety auditor. Such a framework could ensure that excessive scaling remains reserved for accountable actors.
The goal would be to design controls that target only frontier AI development, without impacting smaller-scale AI or consumer electronics. Preliminary ideas include restricting bulk sales of chips below a certain nm threshold and regulating based on end-use. The licensing framework would need detailed planning — such as identifying a safety auditor, defining standards and developing mechanisms against circumvention. Ideally, this framework could serve as an enforcement mechanism for an international AI Auditing Framework and prevent EU-made dual-use machines from contributing to harmful outcomes.
It is important to note that this strategy is quite the leap and likely to meet a lot of opposition. A key issue is that the U.S. is not manufacturing most of their advanced AI chips themselves - they are currently outsourcing most of the fabrication to TSMC in Taiwan. This means that export controls towards the U.S. would not be enough to meaningfully interfere with their access to chips produced with ASML’s technology. The enforcement of this strategy therefore relies on the credible threat that the EU could restrict EUV exports to Taiwan, which could significantly strain relations with them. Furthermore, the U.S. would likely retaliate and might not be easily coerced into negotiations. One way to make this strategy less hostile towards the U.S. would be to include them in the standard setting process and come up with a minimum consensus: potentially trading less strict export controls against stricter safety standards. While this strategy would incur high political and economic costs, a high cost might be acceptable in scenarios where the spillover risks from unsafe U.S. AI development are sufficiently high.
Still, the effects of a strategy like this remain largely unclear and should be further investigated. Further study is for instance required to assess the extent of ASML’s leverage and the likelihood of U.S. cooperation in the face of controls on their technology.
3.1.2 General Considerations
Make it European
European export controls offer advantages over national measures. They distribute countermeasures across the EU, reduce vulnerability to external pressure, and establish the EU as a meaningful geopolitical actor. EU-level controls require only qualified majority voting, potentially bypassing Dutch opposition. A good example of EU export controls are those on personal protective equipment during the COVID-19 pandemic. After member states initially imposed their own restrictions, the Commission replaced them with an EU-wide export authorization scheme, aligning controls with collective interests. The Commission acted through Article 5 of its common export Regulation, which allows for emergency export restrictions in a "critical situation" regarding essential goods. While the EU mainly harmonized existing restrictions, it still overrode national decision-making on export controls through majority voting in the Council. While Dutch controls have the potential benefit of speed, the Netherlands is more vulnerable to retaliation and has large economic incentives not to impose controls. The key uncertainty here is that export controls are still primarily seen as a national competence and might be politically difficult to impose for the EU.
Target and Scale Controls
To maximise impact while minimising costs, any type of export controls should target only the most advanced EUV lithography machines rather than all ASML exports. ASML produces two types of advanced lithography machines - EUV (Extreme Ultraviolet) and DUV (Deep Ultraviolet) lithography machines. EUV machines are used to manufacture cutting-edge chips at 7nm and below. They are ASML’s most advanced and expensive products and generated around 38% of its revenue in 2024. In contrast, DUV machines remain widely used for larger-node chips (e.g., 28nm, 14nm, 7nm), which power everything from cars to industrial equipment. To strike a balance between leverage and economic stability, export controls could start by targeting the newest generation of EUV lithography machines (High NA systems, roughly 2% of ASML’s revenue in 2024). They could later expand to include their TWINSCAN NXE:3800E and TWINSCAN NXE:3600D machines, which are capable of producing 3nm chips and below. Retaining the threat to further scale the controls might incentivise the U.S. to keep retaliations in proportion.
It is important to note that export controls on EUV machines would not affect any already existing chip fabrication capacity in manufacturing countries such as Taiwan, South Korea or the U.S. Instead, they would interfere with efforts to further scale chip production and replace machines, which would affect the ability of chip foundries to keep up with the rising demand. For context, leading AI chips typically remain relevant for frontier model training for less than 5 years before being replaced by newer hardware.
For a more direct impact, it might be more powerful to introduce export controls on ASML’s maintenance services. Maintenance services represent around 20% of ASML’s revenue and are important given the 10-30 year lifespan of ASML machines. The impact of stopping servicing requires further study, though it must be noted that the U.S. has previously applied pressure to restrict ASML from servicing certain Chinese companies, which suggests strategic value. ASML is also speculated to have a ‘kill switch’, initially developed in the context of a potential Taiwan invasion.
3.2. Export Tax / Royalty System
Beyond export controls, the EU could leverage ASML if it wants to ensure it captures a fair share of the economic value generated by AI systems that were built using European technologies. Output-based royalty systems and export taxes are sketches of ideas that could fit in the long-term economic security strategy of the EU, complementing export controls.
The EU could implement an output-based royalty system where chip producers using European technology pay fees based on end applications. This would follow a sliding scale where general computing has minimal fees while AI applications pay progressive fees based on scale and commercial value. ARM, which designs the processor technology found in virtually all smartphones, employs a tiered licensing model where companies pay different rates depending on whether they license just the basic architecture or complete ready-to-use designs. Similarly, nations like Norway and Canada collect oil and gas royalties using sophisticated sliding scales where rates automatically increase as production becomes more profitable — capturing greater value for the public as commercial success grows. This might ultimately require a fundamental shift in ASML's business model, such as to leasing or service-based — allowing the technology provider to capture ongoing value as their equipment generates downstream economic benefits. Details of implementation require further study, especially on who implements and thus captures this windfall — ASML, the EU or the Netherlands — but the system would require dynamic adjustment of royalty rates based on market analysis. Further research is also needed to determine how much downstream value a royalty-based model could realistically capture. ASML’s upstream position limits direct access to end-users like AI developers, who will likely accrue most of the gains. While dynamic pricing for customers like TSMC could push costs downstream, it’s uncertain how much value can be extracted without undermining supply chain efficiency.
3.3. Anti-Coercion Instrument (ACI)
The EU's Anti-Coercion Instrument (ACI) provides an existing framework to counter economic coercion. When a foreign nation attempts to coerce a Member State, such as by imposing punitive trade measures to force specific policy changes, the Commission initiates an examination process, followed by a (qualified majority) Council determination of economic coercion on a 10-week deadline. After attempted diplomatic resolution, the Commission can deploy proportionate countermeasures across multiple trade domains, including export controls. Applied to ASML, given an instance of economic coercion, the ACI could enable restrictions on new exports of EUV machines to the U.S. and potentially on maintenance services for existing equipment, though this warrants further study. While this would not limit their primary supply of chips, which they typically manufacture through TSMC in Taiwan, it would likely interfere with their efforts to scale up domestic advanced chip fabrication.
It must be noted that the ACI is recent and has not been activated yet. The idea has been floated by Ursula Von der Leyen as a retaliation in the ongoing trade war, potentially focusing on tariffs targeting Big Tech Services. Focusing on Big Tech tariffs rather than ASML export controls would likely be more proportional. While the ACI seems like the most readily available tool for export controls on ASML, its narrow focus on reacting to economic coercion makes it difficult to wield strategically. It also remains unclear under which circumstances imposing export controls on ASML would be the most effective measure to guard against economic coercion. It likely is an unnecessarily disruptive tool in the face of economic coercion and might be better suited as a narrow tool for AI policies and agreements such as the ones mentioned above.
4 Policy Challenges & Key Uncertainties
The EU is not a Unified Actor
The EU faces significant hurdles in implementing effective export controls due to its structure as a complex ecosystem rather than a unified entity. Decision-making involves lengthy negotiations at the European Council, where competing national interests often result in watered-down compromises. Despite an EU dual-use regulation, export controls remain primarily a national responsibility, with limited EU oversight. This creates structural barriers to quick, effective and coordinated action. For instance, the Pegasus controversy highlighted loopholes, with ostensibly banned spyware still receiving export licenses.
Economic Impact on ASML and the Netherlands
Any export controls would not only impact ASML, but also the Netherlands and the broader EU economy. ASML is important to the Netherlands' economy, with around half of its 44,000 employees based there. Beyond direct employment, ASML spent €16.0 billion on suppliers in 2024, largely within the EU and the Netherlands. With net sales of €28.3 billion in 2024 distributed globally across China (€10.2B), South Korea (€6.4B), the US (€4.5B), and Taiwan (€4.3B), ASML is extremely vulnerable to export controls. EUV represents 38% of ASML’s revenue, with the newest generation of EUV (High NA) representing 2%. This reinforces the importance of targeting any measures.
A related risk would be ASML moving away from the Netherlands, as was threatened in March 2024. Imposing export controls could strengthen their incentive to leave. However, ASML has enormous physical and human capital investments in the Netherlands. As they did in 2024, the Netherlands would do everything to prevent ASML from leaving.
ASML’s Dependence on U.S. Tech Subsidiaries
Further, ASML faces strong US export control constraints (FDPR) due to its American tech dependence. Its EUV technology license originally came from a U.S. subsidiary, and about 10% of its technology is American (Focus, chapter 29). With more than 8,000 US-based employees through its Cymer laser division and two U.S. board members, reducing ASML's U.S. tech dependence is extremely difficult. This U.S. entanglement has already forced ASML to comply with an increasing number of US-originated and US-pressured Dutch export controls, along with extralegal US pressure and could theoretically impose further export controls. The commercial impact is substantial: ASML has not been able to sell top-end EUV machines to China since 2020, and top-end DUV machines since 2024. While these restrictions were intended to reduce risks — particularly regarding the development of Chinese AI chips and their implications for the global AI race — they also show the extent of U.S. power affecting decisionmaking in the EU. The geopolitical might of the U.S. must not be underestimated, as highlighted by the Ukraine war, tariffs, the (now repealed) AI Diffusion framework or the recent pushbacks against EU Big Tech regulations.
Retaliations
If the EU imposed export controls on ASML's EUV lithography machines, it could trigger major U.S. retaliations. Dutch firms might face sanctions or be blacklisted under the U.S. Entity List. The U.S. could also pressure the EU by imposing further tariffs or threatening trade agreements and its support for NATO. Furthermore, it could leverage export controls on AI chips, essentially cutting the EU off from the advanced chip market — or it could pressure Taiwan to restrict its semiconductor exports to the EU, causing a semiconductor shortage that could affect the EU’s automotive, consumer electronics and medical sectors. While the harms from U.S. retaliation would likely be significant, the costs might be acceptable in scenarios where the spillover risks from unsafe U.S. AI development or the gains from capturing downstream AI value are sufficiently high.
Pitfalls of a Concentrated Supply Chain
While ASML's monopoly provides leverage, interdependence and supply chain concentration further downstream complicates targeted export controls. TSMC holds about 90% market share for advanced chip fabrication, with Samsung and Intel sharing the remaining 10%. The U.S. has been working on their domestic chip fabrication capacities through initiatives such as the TSMC fabrication facility in Arizona, but as of now, Taiwan is the only country able to produce the most advanced chips at scale. As a result, the U.S. outsources their chip fabrication to Taiwan, which limits the impact of direct export controls towards the U.S. If export controls were to include other countries like Taiwan, they would be more effective, but cause significant collateral damage and — unless compensated — threaten partnerships that are essential in the EU’s pursuit of technological sovereignty.
Limited Timeline
ASML's monopoly on EUV lithography machines may have a time limit. Export controls could accelerate U.S. and Chinese efforts to develop domestic manufacturing equipment, eventually reducing EU leverage. However, EUV lithography machines are incredibly complex; Nikon and Canon poured billions of dollars into developing them and failed (Focus, chapter 18). Except for SMEE in China, no one is attempting to develop EUV machines. Shifts in the AI paradigm, such as the growing importance of inference scaling and memory chips built with larger nodes, could also influence the importance of ASML.
5 Potential Mitigation Strategies: Developing Economic Security Capacity
The EU's ability to develop independent export control strategies depends significantly on reducing its technological, economic, and security dependencies. This evolution could be supported through several complementary approaches:
Building Technological Sovereignty and Institutional Capacity
Continued investment in initiatives like the EU Chips Act, and the European Defense Fund strengthens the EU’s technological sovereignty and reduces its security dependence, while diversifying trade partnerships through agreements like the EU-Mercosur Free Trade Agreement and the ongoing free trade agreement negotiations with India might reduce vulnerability to disruptions. The EU currently faces significant personnel constraints in the economic security domain — the European Commission's economic security teams comprise approximately 50-100 staff scattered across the Directorate-General for Trade and Economic Security and DG Connect. In comparison, CHIPS for America in US Commerce had more than 140 people by August 2023. Similarly, the fragmented funding landscape for chips – spanning a panoply of instruments, from the ERC to the EIC, passing by the EIT and EIF, but also InvestEU, Chips JU and SNS JU — suggests potential benefits from greater coordination around strategic priorities. While achieving full independence is a distant goal, each step gradually increases the EU's leverage and reduces vulnerability to retaliations.
Developing a Long-Term Framework
A long-term European economic security approach might draw lessons from structures like the (now repealed) U.S. AI Diffusion Framework, using calibrated thresholds to balance limited short-term costs with long-term controls. Effective implementation would likely require a centralised instrument with clear, dynamic parameters and meaningful EU-level enforcement capacity — elements that warrant further exploration but exceed this brief's scope.
Mitigating Costs: Semiconductor Security Fund
Even with greater coordination and independence, an EU export control strategy would be costly. A European Fund for Semiconductor Security (EFSS) could provide targeted support to stakeholders affected by export controls — particularly ASML and the Netherlands — while serving as a safeguard against other supply chain disruptions. This approach would enable the EU to share the burden of export controls and prevent critical industries from being crippled by sudden supply shortages. Details of how this would be funded are relevant but out of scope for this brief.
6 Conclusion
ASML’s monopoly on lithography machines provides a unique opportunity for the EU to limit catastrophic AI risks and advance long-term EU economic security. Expanding EU oversight in export controls presents political challenges, potentially requiring shifts in the Overton window regarding an increased awareness about AI risks and stakes as well as granting the EU more power. Nevertheless, this analysis suggests there may be strategic value in exploring such options. A centralised framework regarding ASML's position in the semiconductor supply chain could potentially offer the EU additional tools in navigating an increasingly multipolar world.
In line with the goals outlined in this brief — first, avoiding spill-over costs from unsafe frontier AI development, second, capturing a sufficient portion of the value AI creates and third, fighting against economic coercion — we have highlighted three venues for further research. These include AI safety licensing requirements, an export tax or royalty system and European-level export controls on advanced EUV lithography machines via the Anti-Coercion Instrument. We conclude that ASML as a strategic asset is likely better used for narrow AI agreements and policies and is less suitable as a blunt economic tool.
The six implementation challenges — from the EU's fragmented decision-making structure to ASML's U.S. tech dependencies and potential retaliations — suggest that any approach would likely benefit from broader (economic) security considerations. These might include investments in technological sovereignty through initiatives like the EU Chips Act, expanded capacity for economic security analysis, and improved coordination across funding instruments. However, even with supportive measures, leveraging ASML is likely a costly strategy which may only be viable in scenarios where the stakes are sufficiently high. Any of the strategies we highlighted rely on a certain amount of speculation about the extent of ASML’s leverage, the costs involved in implementing these strategies and the ripple effects that might be caused in the process.
Many questions remain unanswered, but a European approach might offer potential benefits in terms of negotiating position, consistent application, and distributed impact. Yet, one must recognise that the joint might of the EU is not that of the US or China, and so these measures need to be realistic. ASML as a bargaining chip is severely understudied, and this brief serves to open up this discussion and invite further research.
